Artificial Intelligence: Is it”Smarter” than us

In class we’ve discussed in varying degrees whether or not Artificial Intelligence is already “smarter” than us. Calculators and Microsoft Excel Sheets can perform mathematical equations faster than humans could dream of solving them. Computers have been beating World Chess Champions since Deep Blue beat Kasparov in 1997 and have only been making improvements since. However, when it comes to what we value as “True Intelligence”, the stuff that helps us define our basis of human nature, it seems at times that there’s no way that a device like Siri or Google Home could be “smarter” than you.

irobot

What a “True Intelligence” Robot could look like. iRobot

Computers are programmable machines which can process massive amounts of information and perform logical equations using that data. They dominate in the quantitative space because according to Time Magazine, they are “not affected or influenced by emotions, feelings, wants, needs and other factors that often cloud the judgement and intelligence of us mere mortals”. Their pure computing power allows them to overcome natural human issues such as fatigue or memory capacity. Speed is where computers like IBM Watson excel. Watson diagnosed a patient with a rare form of Leukemia in less than 10 minutes using genetic data which normally would’ve taken 2 weeks by human experts. According to the leading researcher on the team, Watson did not catch something that wouldn’t have been caught by the doctors but the speed with which it was caught prevented complications and issues that arise quickly with forms of leukemia.

giphy17

IBM Watson

But what is true intelligence? The Jerusalem Post recognizes that people often misconstrue what artificial intelligence is and what some of the jargon used to describe these “smart” machines mean. Artificial Intelligence is the “ability to solve problems and learn”. Shlomo Maital elaborates, “Learning and problem solving are related. The more problems you solve, the more you learn. And the more you learn, the better you get at solving problems”. So if computers are able to learn from the previous information we feed it AND are able to discern new information each time it solves a problem, why don’t we feel that AI is taking over the world?

tumblr_inline_nedpw5obbp1t2m9ry

Machine learning is a subset of Artificial Intelligence which says that a set of computer algorithms will get better and better at solving issues with each solution it creates. For example, AI has been able to use camera footage to predict where Hamas rocket launchers will be set up and can name suspicious objects in camera footage. This information is relayed to a military unit for action to be taken. Another example is Waycare’s platform which uses machine learning and deep learning to predict traffic accidents about 2 hours before they occur. This allows police units to be dispatched to the area to reduce response time or to prevent accidents from happening. Both of these examples show that computers are able to predict things that may happen. These machines clearly are better at predicting when specific events may occur but still don’t cross that “cool/creepy” line.

waycare.png

Artificial Intelligence struggles with context and the “general knowledge” component of Intelligence. Phil Wainewright presents a great example of where artificial intelligence recognition sometimes ends up comparing apples to oranges and has trouble discerning when it’s doing this without specific instruction. Google, Amazon and Microsoft are able to digitally recognize and tag images with general tags like “ocean”, “nature”, and “water” but a company like Booking.com actually needs to know when looking at an image if there’s a “sea view” or if the room has a “balcony with a seating area”. While these services might be able to recognize objects individually, the machines originally weren’t able to recognize the context of the objects that were incorporated, something that most young/teenage adults could do easily. The article goes on to explain that it is the concept of new category creation and spontaneous knowledge building which claims to be the distinguishing factor of why Wainwright thinks that Humans will always be smarter than A.I.

google-home-product-photos-27.jpg

Currently, Google Home and Amazon’s Alexa are flooding the country with social robots, able to interact as a physical representation of a digital friend. With different emphases, these robots aim to act as a digital personal assistant and complete certain tasks. While Siri seems to never want to cooperate with iPhone users, except when trying to text a friend or skip to the next song, some of these in home voice assistants have made some major strides in their conversation recognition and logic. Google home is the device which is leading the space in standalone quality and tends to be smarter than Amazon Echo according to Andrew Gebhart. Google Home is able to understand context of questions, something which Amazon has been playing catch up on recently. Google Home also is backed by none other than Google Search, an infinite treasure trove of information that needs to parsed by Google to be delivered in a conversational manner. Google also has voice recognition in its device, giving it the ability to deliver a catered answer to questions like “call my mom” or “what’s my calendar like today” rather than having to ask whom the device is speaking to…or avoiding accidentally calling your mother in law and not your mother (a hilarious example presented). I sure hope that the Amazon Echo doesn’t squash the development of a better base technology (Google Home) because it has better compatibility (cough…cough…VHS over BetaMax). Nevertheless, these devices are attempting to replicate what it feels like to have an assistant but their functionality is limited to the skills that are built. Unfortunately, having a conversation with an Amazon Alexa doesn’t feel conversational after the first time novelty wears off. The “social nature” of these devices still comes off as functional rather than dynamic and spontaneous.

unfortunateinsecureballpython-size_restricted

Jibo, A Social Robot

I think lastly I would like to mention Jibo, a social robot whom I thought would change the way that the massive tech giants Google and Amazon thought about their smart speakers. Jibo boasted in its original Kickstarter that the little robot who stands about the height of a small lamp would be able to make video calls and read bedtime stories. Jibo appears to be a cute friend that sits on a nearby counter and a quick glance at a demo reveals just how lifelike and smooth his 360 Degree swivel movements are. Unfortunately, where the realism falls short is in Jibo’s functionality. Jibo is able to perform minor social tasks like telling you a joke, fun fact and even recognize you. However, Jibo is not able to answer your pondering questions or connect to calendars the same way that Google Home is able to. The connectivity and ecosystem of products is lacking for Jibo, so its knowledge is limited. Therefore, the promise of a truly social and fun robot to have around the house, the perfect roommmate, will just have to wait.

Automaton

As we combine the ideas that Artificial Intelligence and Machine Learning can complete tasks better than we can with a social emphasis like Google Home and Amazon Alexa, maybe someday we will have a robot that “appears” to be truly smarter than us. But until robots “contextual awareness” and knowledge of infinite pattern recognition can improve, we will continue to create robots that feel like glorified automatons working towards tasks and not the recreation of iRobot.

 

 

 

9 comments

  1. Jobabes121 · ·

    Great post, Tucker! You did a great job of explaining machine learning in a very concise manner with specific examples. I did not know that it can even predict traffic accidents in advance… that really caught me off guard as there are simply so many possibilities where certains drivers may not cross one particular way, etc. Something that seems completely random can be predicted and estimated via machine learning and AI development. Machine learning’s capability of predicting such events almost seemed like a portfolio’s expected return calculation, where each security (event) would yield a random payoff yet data analysis (this case, machine learning with repeated information) allows us to predict the portfolio’s overall expected return. Sounds incredible to me, and I am excited to see more to come in the future.

    In regards to our superiority over AI, I completely agree. Whenever I use Alexa, it does not have any contextual awareness that you mentioned in this blog. Unless it is instructed or coded to recognize a particular setting from a specific language or sentence, it will never be able to capture it. Despite the fact that AIs can act in a way that we do not understand with constant churning of data, humans will always have a better macro-level understanding of decision making, which is always necessary in any business setting.

  2. kylepdonley · ·

    The concept of new category creation is something I wasn’t previously aware of with AI. The ability to learn quickly and with infinitely better recall than humans, AI always seemed like it would be a better “learner” than humans. But knowing this is one of the big stumbling blocks is kind of an ah-ha moment for me in that the human brain will always be more powerful in that it is not restrained. Very interesting post, thanks for sharing!

  3. kikinitwithraf · ·

    Great post @tuckercharette. And while machine learning and AI will always perform at faster rates, the one thing they fail at is the emotional connection that humans bring to the table. Similar to the Jibo joke example, there will always be a way to program “machines”, but will they be able to think for themselves in a construct not based on commands? Thats where it all gets interesting.

  4. phanauer1 · ·

    The question of whether or not AI is smarter than us is complicated and subjective, but the way that I would boil it down is that AI has greater capacity for learning than us and more potential, so It could become smarter than us, but it is not yet. As you mentioned, I think the really impactful strides in this industry are yet to come, and that’s when we might start to see technology that is really smart and has “true intelligence”. Until machines can better understand context and human thought processes, they won’t be able to work well with us and we will remain one step ahead; after that, who really knows… great post!

  5. Tully Horne · ·

    I think you had great examples in this blog that highlight the importance and effectiveness of the current uses of AI. I think the best way to describe it is that as long as AI remains an extension of ourselves and doesn’t become completely independent it won’t get out of control. If we can harness its power while maximizing its utility the possibilities with it are endless. I do think AI will be smarter than us one day, but it first has to learn many things, like what you discussed as well as understanding sarcasm, tone, etc. across different languages. It will be interesting to look back in 5/10 years on this class and our discussions on this topic and see how accurate we were and how our views have changed.

  6. roarkword · ·

    You made a good point about the interconnectivity of the ecosystem needing to be more robust if a device like Jibo are going to succeed. In that same vein, it seems to me that Jibo will need to be acquired by Google if they hope to achieve the same functionalities that Google Home has. But even still if the two are merged, they will still lack that crucial contextual awareness that you pointed out; this awareness will be where the real line of creepy and cool get toed. I suspect that as people become more taken aback by the ability of an AI to interpret the subtleties of what only a human had previously been able to do, then people will be both blown away and carefully suspicious. All that being said, I totally agree with @tullyhornebc and what looking back on our musings will be like as these advancements actually take place!

  7. Nice post. I just sent you an email about Wargames. Just watched it with my kids, it had some interesting connotations for AI (i.e. optimizing on the wrong outcome).

  8. JohnWalshFilms · ·

    Great post, Tucker, and a great overview of the technologies dominating this AI space, as well as the importance of improving AI’s ability to “think” in context. In your research, have you found that the Google product has been “smarter” due to the huge search engine that backs its machine learning? How does Google better equip its “Google Home” device to think in context – is it a coding and programming endeavor, or a specific methodology to the “machine learning” approach?

    1. tuckercharette · ·

      Hey John, just to address your question. It seemed that Google was able to more seamlessly integrate its answers into voice responses from the Google Engine rather than merely piping back results from Google in a web page form like Siri seems to do. I’m not quite sure whether its a coding and parsing issue or if it’s a machine learning approach.